Predictive Analytics and College Completion: Overcoming Academic and Financial Obstacles

Despite the fact that 70% of recent high school completers enrolled in college in 2016, the number of them graduating from college has not kept pace.¹ Only 36% of those aged 25 to 29 in 2016 had a bachelor’s degree or higher.² In addition, while racial gaps in degree completion have narrowed, these gaps have not narrowed nearly enough, with 22.7% of Black adults and 18.5% of Hispanic adults aged 25 to 29 having completed a bachelor’s degree, versus 43% of those who are white and 60% of those who are Asian.²

Indeed, the student groups historically disenfranchised from college – low-income, Black, Latino/a, and academically at-risk students – continue to struggle to college completion. This is especially true for those who experience financial obstacles to college access and completion. This population experienced increasing enrollment rates but stagnating completion rates.^{3 4} And even when academic achievement is considered, lack of financial resources is still a significant obstacle to college completion.⁵ In the case of one state institution, 1/3 of students who left school had 3.0 or higher GPAs,⁶ so academic achievement alone is not enough to get students to college completion. What was previously an issue of enrolling more students in college is now an issue of too many students starting college but not completing it.⁷ This leaves thousands of students with college debt but no college degree.

Luckily, with advances in technology and increasing access to information regarding student persistence and retention trends, Early Warning Systems (EWS) that use predictive analytics allow colleges and universities to efficiently connect with students, assist them in developing a plan for success, intervene when necessary and in the right way, and track their progress. Predictive analytics anticipates issues before they happen and flag at-risk students to queue intervention.

The power of predictive analytics is in its ability to leverage known data about the student to not just determine which students need assistance but also to determine what challenges a student might face on their path to success. To do so these methods use data harvested from campuswide systems that can range from basic ‘geographic’ and ‘demographic’ features to more actionable ‘academic ability’, ‘performance’ and ‘behavioral’ features on campus that can indicate when a student might be falling off path. In addition, these systems may or may not use the important, yet sensitive, ‘ability to pay’ and ‘financial support’ insights gathered from the Free Application for Federal Student Aid (FAFSA) and financial aid system on campus. These specific data points can often further explain why some students may experience roadblocks while looking to complete their degree objectives. While data gathered from the FAFSA form submitted by students is commonly leveraged in the early stages of a relationship between a student and an institution (i.e during financial aid packaging), this paper outlines the impact this information can have on an institution’s ability to meet the needs of students during later stages of that relationship, specifically once a student is enrolled and actively pursuing their degree. We present a brief literature review regarding data found to be impactful on determining the risk associated with student success, in addition to documented concerns over how these data points can be used under recently published interpretations of the Higher Education Act and FERPA. Ultimately the purpose of this paper is to encourage a discussion on how we can best serve students facing financial barriers, especially the low-income, minority, and academically at-risk students most prone to drop out. Considering financial barriers not just during the admission and enrollment phase but also during the academic year to refine student aid alongside academic guidance may provide a clearer path to student success.

The FAFSA is the form used by Federal Student Aid (a part of the U.S. Department of Education) to allocate more than $120 billion in federal aid each year, and is used to make decisions regarding financial aid awards at the state and college/university levels.^{8 9} While some student-level data can be gleaned from the student’s application to the institution, other information is key to understanding the student’s socioeconomic background, and its impact on college outcomes can best be gathered from the FAFSA form. First, this information includes family characteristics, as measured by the marital status of the student and/or parent(s), and whether the student is a dependent and/or has any dependents. Second, income is included as measured with the student’s and/or parent’s adjusted gross income and the expected family contribution (EFC). This information in turn provides institutions with the ability to determine a student’s documented need respective to the institution’s costs and ultimately an understanding of needs met during financial aid processing. Below, we outline the role that each of these data points can play in understanding college retention and completion.

For both traditional and nontraditional students, family structure is important to consider when tracking college retention patterns. For instance, seeing the increase in the number of nontraditional students in college and universities increase in the last 20 years,¹⁰ understanding whether or not the student is married is important. For instance, in a study of firsttime, full-time, degree-seeking students, the authors found that current marital status was a significant predictor of stopout and dropout behaviors, with marriage status having differential effects on whether men and women stopout, dropout, or continuously enroll.¹¹ A different study of Hispanic women found that persistence in college was related to whether or not these women delayed marriage and childbirth.¹² It is possible that these students need more support, advice, and guidance from their academic advisors on balancing academics, costs, and home life. Information on family characteristics, thus, can help colleges and universities target those students who may need additional academic, personal, and financial support.

Having dependents is also a significant predictor of stopout and dropout for students,¹¹ a reality for upwards of 25% of college students today.¹⁰

Family income is one of the most significant predictors of college success, and the inability to pay for college is one of the foremost reasons that low-income students drop out of college.¹³ Using the student’s income, the parent’s income, and most importantly the EFC, those interested in predicting college student success can measure need-based aid eligibility. From this research, we know that students from lower-income households and those eligible for need-based aid are less likely to graduate from college.^{14 15} With the help of academic advisors and those using EWS to flag students who are financially in trouble of dropping out or stopping out, this financial information can be used as a talking point for advisors to help students determine if they need more financial aid, and refer students to the financial aid office.

Because of the financial obstacles many students face, college completion for many students hinges on whether or not they receive enough financial support to afford college.^{16 17} Moreover, including information on the amount of aid awarded in addition to EFC is crucial because these two variables have opposite effects: lower-income students (who have a lower EFC) are less likely to graduate, but those whose need is met are more likely to graduate. Therefore, both pieces of information are necessary to ensure accurate predictions of persistence to graduation.^{14 15} This is especially true for Black and Latino/a students; much of the race gap in graduation (at least at elite institutions) can be remedied with financial aid.¹⁴ Black and Latino/a students are also more sensitive to the receipt of financial aid than white students.¹⁸ Sensitivity to aid also differs by income. Low-income students are more sensitive to the amount of aid given, which means that failing to include this information will disproportionately harm low-income students.⁵

The type of aid is also important to consider, as loans, grants, and work-study monies all exert different effects on persistence.

Alon (2007) found that minority student persistence was particularly sensitive to grants and scholarships, whereas loans and work-study funds were less helpful in persisting, a finding with support elsewhere.^{5 14 19} In a study regarding the separate issues of stopout and dropout, researchers surmised that loans (versus no aid) may actually increase dropout, while work-study aid may decrease dropout.¹¹

Understanding each of these characteristics specific to students on campus can certainly provide academic advisors an opportunity to talk about financial aid obstacles to college completion and in turn direct students to resources that can help overcome these potential risks to success.

Community college students are a group particularly vulnerable to attrition, with nearly half not persisting to their second Fall semester.²⁰ Community college students are increasingly nontraditional, which means they are older than the average college student, are more likely to be Black or Latino, are more likely to come from low-income households, and are more likely to be underprepared.^{8 21} Thus, they are more susceptible to some of the concerns mentioned above, and therefore great potential exists to support these students with Early Warning Systems that can use financial information. For instance, even beyond income or EFC information, just having information on whether students completed the FAFSA (at times separate from whether funding was awarded) can be key to understanding persistence and completion rates, especially part-time students.⁸ Academic advisors can use this information to encourage students to complete the FAFSA – even at a late stage – or in preparation for the following academic year.

With an understanding that income and financial aid information can be vital to supporting college students and reducing dropping out and stopping out behaviors, there is certainly confusion and concern about how this information can be used. Recently the National Association of Student Financial Aid Administrators (NASFAA) has distilled guidance based on their interpretation of the Higher Education Act and has in turn published a decision tree to help institutions determine how to share and use this information.²² Sharing financial information (with offices within or outside the institution) could fit under provisions 3a and 10 of this decision tree. Since the information crucial to determining students at risk of dropping out due to financial obstacles comes from the FAFSA, this information, per 3a, needs to “be used for the application, award, and/or administration of Title IV funds, state aid, and/or institutional aid programs under HEA 483(a)(3)(E)?” In addition, point 10 indicates disclosure is permitted in cases where “the disclosure to an organization conducting research for, or on behalf of, your institution to: (A) Develop, validate, or administer predictive tests; (B) Administer student aid programs; or (C) Improve instruction.” In other words, this information can be shared with entities (both on and off campus) if it is used to administer student aid.

This last point is an important one as it points to the overlap in the role of academic advisors to help students stay on track and address obstacles to academic success. Per NACADA, the professional organization of academic advisors, advisors can and often do examine financial considerations when discussing academic plans with students.²³

For instance, advisors may need to consult with the financial aid office when ensuring students take the right amount of classes to match their financial aid.²³

Advisors particularly help students who may be at risk of not meeting Satisfactory Academic Progress (SAP) to maintain their federal aid.^{24 25} Even more importantly, if an advisor sees that a student is struggling to pay for college, or is hitting time limitations associated with loan eligibility and the Pell Grant, they can and should find additional financial support for the students in conjunction with the financial aid office. Advisors can also take a long view, discussing degree completion and how a student plans to pay for college over the length of the degree.

However, this individual student outreach can be difficult to do in institutions with high student-toadvisor ratios, and patterns across students are not always clear. But when paired with dedicated advisors, predictive analytics using academic and financial aid considerations can increase the number of students meeting Satisfactory Academic Progress, and increase the overall retention of the student body. For instance, Early Warning Systems and processes have already found success at major colleges and universities. The University of Kentucky has paired their predictive models with decisions regarding financial aid, providing additional assistance to students most likely to drop out due to financial barriers.⁶ Their efforts have led to a 10-20% increase in students returning to the school who would have otherwise left. The University of Central Florida used predictive analytics including financial information to discover a large number of students in one residence hall who were dropping out because of financial reasons, and includes 16 financial risk factors to its EWS to give academic advisors this information.²⁶

As colleges and universities continue to focus on increasing retention and completion rates so as to ensure financial vitality and high-quality academic programs,²¹ the increasing use of data can help them meet these goals, and importantly, help students overcome the obstacles to a college degree. Knowing how devastating the lack of financial aid or the ability to pay for college can be for graduation rates,^{5 7} we must use our best efforts to flag these and other obstacles to college completion. Predictive analytics have been used to close the retention gap and alert advisors and other end users of students at risk of dropping out.²⁶ This technology has the power to assist those students who have traditionally struggled in degree completion, closing the racial and income gaps in college completion and decreasing overall economic inequality as the education gap is leveled. Even academically qualified students will find financial obstacles prevent them from graduating with a college degree,⁶ so utilizing financial in addition to academic information is imperative to supporting all students.